Defect classification device and defect classification program

Abstract

An object of the present disclosure is to provide a defect classification device capable of easily grasping an appropriate recipe update timing of an imaging device when classification accuracy for classifying defects existing on a semiconductor wafer is decreased. The defect classification device according to the present disclosure calculates classification accuracy by further acquiring a result of a manual classification for defects spanning a plurality of classification spaces as a result of an automatic classification, and comparing the result of the automatic classification with the result of the manual classification (see FIG. 5 ).

The invention claimed is: 
     
         1 . A defect classification device configured to classify at least one defect on a wafer, the device comprising:
 an image acquisition unit configured to acquire an observation image from an imaging device configured to capture the observation image of the wafer;   a computer system configured to classify the at least one defect in the observation image; and   a user interface configured to input an operation instruction for manually classifying the at least one defect in the observation image, wherein   the computer system is configured to   classify the at least one defect in the observation image into one of
 a first feature space, 
 a second feature space different from the first feature space, and 
 a third feature space spanning both the first feature space and the second feature space, 
   further receive a result of a manual classification via the user interface for each defect classified into the third feature space among the at least one defect, and   calculate and output an index indicating classification accuracy of the at least one defect by using a result of classifying the at least one defect into the third feature space by the computer system and the result of the manual classification,   wherein, when the classification accuracy is less than a threshold value, the computer system is configured to change an imaging condition used when the imaging device acquires the observation image or to output an alert for prompting the change, and   wherein two or more classification classes are provided in an unknown class, including a first unknown class corresponding to a region closer to a boundary between the two or more classification classes and a second unknown class corresponding to a region farther from the boundary, and a number of defects belonging to the second unknown class is used to calculate the index.   
     
     
         2 . The defect classification device according to  claim 1 , wherein the computer system is configured to
 perform a first classification step of classifying the at least one defect into one of the first feature space and the second feature space,   perform a second classification step of classifying the at least one defect into one of the first feature space, the second feature space, and the third feature space,   extract, from a classification result in the first classification step, each defect classified into the third feature space in the second classification step as an evaluation target defect to be used for classification accuracy determination, and   determine the classification accuracy by using a number of defects classified into the third feature space in the second classification step and the number of defects classified into the third feature space by the manual classification.   
     
     
         3 . The defect classification device according to  claim 2 , wherein the computer system is configured to
 acquire, a number of correct defects in which a classification result by the manual classification matches the classification result by the first classification step among the defects classified into the third feature space in the second classification step, and   output, as the classification accuracy, a value obtained by dividing the number of the correct defects by the number of the defects classified into the third feature space in the second classification step.   
     
     
         4 . The defect classification device according to  claim 1 , wherein the computer system is configured to
 perform a first classification step of classifying the at least one defect into one of the first feature space and the second feature space, and   perform a second classification step of classifying the at least one defect into one of the first feature space, the second feature space, and the third feature space, and   the user interface includes a correction interface used for correcting a result of the first classification step while comparing a classification result by the first classification step with a classification result by the second classification step.   
     
     
         5 . The defect classification device according to  claim 1 , wherein
 the third feature space includes at least a part of a boundary between the first feature space and the second feature space.   
     
     
         6 . A non-transitory computer-readable medium storing a defect classification program causing a computer to execute a process of classifying at least one defect on a wafer, the process including:
 a step of acquiring an observation image from an imaging device configured to capture the observation image of the wafer;   a step of classifying the at least one defect in the observation image; and   a step of receiving an operation instruction from a user interface for inputting the operation instruction for manually classifying the at least one defect in the observation image,   in the step of classifying the at least one defect, the defect classification program causes the computer to classify the at least one defect in the observation image into one of
 a first feature space, 
 a second feature space different from the first feature space, and 
 a third feature space spanning both the first feature space and the second feature space, and 
   the defect classification program causes the computer to further execute   a step of further receiving a result of a manual classification via the user interface for each defect classified into the third feature space among the at least one defect, and   a step of calculating and outputting an index indicating classification accuracy of the at least one defect by using a result of the step of classifying the at least one defect and the result of the manual classification,   wherein, when the classification accuracy is less than a threshold value, the computer is configured to change an imaging condition used when the imaging device acquires the observation image or to output an alert for prompting the change, and   wherein two or more classification classes are provided in an unknown class, including a first unknown class corresponding to a region closer to a boundary between the two or more classification classes and a second unknown class corresponding to a region farther from the boundary, and a number of defects belonging to the second unknown class is used to calculate the index.